Synthetic data pipeline generates annotated training data for scratch detection

Researchers have developed ScratchSim, a procedural synthetic data pipeline using BlenderProc to generate annotated training data for surface scratch detection. This method addresses the challenge of limited annotated defect data in industrial quality control. The pipeline offers configurable material appearance, camera modes, and domain randomization, producing automatic COCO-format annotations. Evaluations demonstrated that fine-tuning models with synthetic data outperforms real-only training, and mixed training effectively recovers performance with scarce real data, showing promise for on-device industrial inspection. AI

IMPACT Enables scalable defect detection in industrial settings by reducing reliance on large, real-world annotated datasets.

RANK_REASON The cluster describes a research paper detailing a new synthetic data generation pipeline for computer vision tasks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• BlenderProc
• COCO
• Ferrari
• Hugging Face
• LW-DETR
• Saptarshi Neil Sinha
• ScratchSim
• YOLO26
• YOLOX Object Detection

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →